CODEWITHSUNDEEP
csegmentation-faultmatrix-multiplicationgsl
user1492775
user1492775

Reputation: 21

Segmentaion fault in C when doing matrix multiplication

I use gsl random generator to generate 2 big matrices, and use gsl cblas to multiply them, but I always got Segmentation fault when the cblas operation begins. When I can't solve this, I then write the code below, using the very basic idea to do matrix multiplication, and I still got Segmentation Fault, but all the two can work all right when matrix is really a small one, I'm really puzzled about this.

#include <stdlib.h>
#include <stdio.h>
#include <math.h>
#include <time.h>
#define PI 3.1415926

void GenerateKey(int m, int n, int l, int q, float alpha)
{
    // initialization
    int i;
    int j;
    int k;

    float *A;
    float *S;
    float *E;
    float *B;

    float sigma = (alpha * q ) / sqrt(2 * PI);
    A=(float*)malloc(sizeof(float)*(m*n));
    S=(float*)malloc(sizeof(float)*(n*l));
    B=(float*)malloc(sizeof(float)*(m*l));
    E=(float*)malloc(sizeof(float)*(m*l));

    // init A
    for(i = 0; i < m*n; i++)
    {
        A[i]=0;
    }
    printf("\n");

    // init S
    for(i = 0; i < n*l; i++)
    {
        S[i]=0;
    }

    printf("\n");

    // init E
    for(i = 0; i < m*l; i++)
    {
        E[i]=0;
    }

    printf("\n");
    float po;
    for(i = 0; i < m; i++)
    {
        for(j=0; j<l; j++)
        {      
            po=0;
            for(k=0; k<n; k++)
            {
                po +=A[i*m+k]*S[k*n+j];
            }
            po += E[i*m +j];
            B[i*m+j]=((int)po) % q;
        }
    }

    printf("Game over");

    printf("\n");
    free(A);
    free(B);
    free(S);
    free(E);
}

int main()
{
    GenerateKey(2680,191,64,72973,0.000551);

    return 0;
}

Upvotes: 1

Views: 405

Answers (3)

Andriy
Andriy

Reputation: 8594

You incorrectly compute an element index for all the matrices.

When you have an MxN matrix which is allocated as an 1-dimensional array, the index for an element (i,j) is i*N+j. Instead, you're computing it as i*M+j.

Upvotes: 1

hexist
hexist

Reputation: 5240

When you're doing i*m+j, shouldn't that be i*l+j ? Similarly with i*m+k should be i*l+k and k*n+j should be k*l+j

The reason being, take for example E = (float*)malloc(sizeof(float)*(m*l)), so you have m rows and l columns (or vise versa), so if you are iterating over the m dimension you need to be multiplying your m iterator (i in this case) by the stride of your matrix in that dimension, which is l.

Upvotes: 3

unwind
unwind

Reputation: 399753

You don't check the malloc() return values, so my guess is that one or more allocation is failing and you're dereferencing NULL. Another possibility is of course an indexing error, so you acccess out of bounds.

Upvotes: 1

Related Questions